Research Parallel — Hermes Plugin

Parallel web research via self-hosted Firecrawl + Perplexity.

Tool: research_parallel

Breaks a query into sub-angles, searches them simultaneously, optionally calls Perplexity for a search-augmented signal, scrapes top pages, and synthesizes a structured report.

Parameters

Param	Type	Default	Description
query	string	—	Research question (required)
depth	int	3	Sub-queries (1-6)
breadth	int	5	Results per sub-query
perplexity	bool	true	Call sonar-pro-search
scrape	bool	true	Scrape top URLs

Architecture

research_parallel(query, depth=3, breadth=5)
  ├─ Phase 1: LLM Decompose → N sub-queries
  ├─ Phase 2: Parallel SearXNG search via Firecrawl (asyncio)
  ├─ Phase 3: Perplexity sonar-pro-search via LLM proxy (concurrent)
  ├─ Phase 4: Parallel scrape of top URLs (asyncio)
  └─ Phase 5: LLM Synthesize → structured report

Dependencies

• Firecrawl API on 127.0.0.1:3002 (self-hosted)
• LLM Proxy on 127.0.0.1:8646 (OpenAI-compatible)
• httpx (async HTTP)

Enabling

hermes plugins enable research-parallel

Testing

cd ~/.hermes/plugins/research-parallel
python3 -m pytest tests/ -v